AWS Marketplace

Publish algorithm on the AWS Marketplace

Create your algorithm and model package

• Building your own container as Algorithm / Model Package
  • When should I build my own algorithm container?
  • Permissions
  • The example
  • The presentation
  • Part 1: Packaging and Uploading your Algorithm for use with Amazon SageMaker
    • An overview of Docker
    • How Amazon SageMaker runs your Docker container
      • Running your container during training
        • The input
        • The output
      • Running your container during hosting
    • The parts of the sample container
    • The Dockerfile
    • Building and registering the container
    • Testing your algorithm on your local machine or on an Amazon SageMaker notebook instance
  • Part 2: Training, Batch Inference and Hosting your Algorithm in Amazon SageMaker
    • Set up the environment
    • Create the session
    • Upload the data for training
    • Create an estimator and fit the model
    • Batch Transform Job
      • Batch Transform Input Preparation
      • Run Batch Transform
        • Inspect the Batch Transform Output in S3
    • Deploy the model
      • Choose some data and use it for a prediction
      • Cleanup Endpoint
  • Part 3 - Package your resources as an Amazon SageMaker Algorithm
    • Algorithm Definition
      • Training Specification
      • Supported Hyper-parameters
      • Inference Specification
      • Validation Specification
    • Putting it all together
      • Describe the algorithm
  • Part 4 - Package your resources as an Amazon SageMaker ModelPackage
    • Model Package Definition
      • Inference Specification
      • Validation Specification
    • Putting it all together
      • Describe the ModelPackage
    • Debugging Creation Issues
    • List on AWS Marketplace
  • Notebook CI Test Results

Curate your sample notebook

Templates

• Train, tune, and deploy a custom ML model using For Seller to update: Title_of_your_ML Algorithm Algorithm from AWS Marketplace
  • Pre-requisites
  • Contents
  • Usage instructions
  • 1. Subscribe to the algorithm
  • 2. Prepare dataset
    • A. Dataset format expected by the algorithm
    • B. Configure and visualize train and test dataset
    • C. Upload datasets to Amazon S3
  • 3: Train a machine learning model
    • 3.1 Set up environment
    • 3.2 Train a model
  • 4: Deploy model and verify results
    • A. Deploy trained model
    • B. Create input payload
    • C. Perform real-time inference
    • D. Visualize output
    • E. Calculate relevant metrics
    • F. Delete the endpoint
  • 5: Tune your model! (optional)
    • A. Tuning Guidelines
    • B. Define Tuning configuration
    • C. Run a model tuning job
  • 6. Perform Batch inference
  • 7. Clean-up
    • A. Delete the model
    • B. Unsubscribe to the listing (optional)
  • Notebook CI Test Results
• Deploy For Seller to update: Title_of_your_ML Model Model Package from AWS Marketplace
  • Pre-requisites:
  • Contents:
  • Usage instructions
  • 1. Subscribe to the model package
  • 2. Create an endpoint and perform real-time inference
    • A. Create an endpoint
    • B. Create input payload
    • C. Perform real-time inference
    • D. Visualize output
    • E. Delete the endpoint
  • 3. Perform batch inference
  • 4. Clean-up
    • A. Delete the model
    • B. Unsubscribe to the listing (optional)
  • Notebook CI Test Results

Use AWS Data Exchange products

• Using Shutterstock’s Image datasets to train a multi-label image classification model
  • Introduction
  • Prerequisites
    • Step 1: Set permissions and environment variables
  • Data Preparation
    • Step 2: Parse the metadata .csv file to create a .lst file for training
    • Step 3: Create the training and validation datasets
    • Step 4: Copy the training data to your SageMaker S3 Bucket
  • Training our Multi-Label Image Classification Model
    • Step 5: Set the training parameters
    • Step 6: Set the algorithm parameters
    • Step 7: Specify the input data
    • Step 8: Train the model
  • Inference
    • Step 9: Deploy the Model for Inference
    • Step 10: Evaluate the model
  • Clean up
  • Notebook CI Test Results
• Using dataset product from AWS Data Exchange with ML model from AWS Marketplace
  • Contents:
    • Usage instructions
  • Pre-requisites:
    • Pre-requisite 1:
    • Subscribe to data from AWS Data Exchange:
    • Pre-requisite 2:
    • Subscribe and deploy ML Model from AWS Marketplace:
  • Introduction
  • Explore dataset
  • Perform inference
  • Next Steps
  • Cleanup
• Notebook CI Test Results

Use AWS Marketplace algorithms

• AWS Marketplace Product Usage Demonstration - Algorithms
  • Using Algorithm ARN with Amazon SageMaker APIs
  • Overall flow diagram
  • Compatibility
  • Set up the environment
    • Create the session
  • Upload the data for training
  • Creating Training Job using Algorithm ARN
  • Run Training Job
  • Automated Model Tuning (optional)
  • Batch Transform Job
    • Batch Transform Input Preparation
      • Inspect the Batch Transform Output in S3
  • Live Inference Endpoint
    • Choose some data and use it for a prediction
    • Cleanup the endpoint
  • Notebook CI Test Results

Autogluon

• AutoGluon-Tabular in AWS Marketplace
  • Contents:
  • Step 1: Subscribe to AutoML algorithm from AWS Marketplace
  • Step 2 : Set up environment
  • Step 3: Get the data
    • Check the data
  • Step 4: Train a model
  • Step 5: Deploy the model and perform a real-time inference
    • Deploy a remote endpoint
    • Predict on unlabeled test data
    • Predict on data that includes label column
    • Check that classification performance metrics match evaluation printed to endpoint logs as expected
  • Step 6: Use Batch Transform
  • Step 7: Clean-up

AutoML

• Goal: Train a model using AutoML functionality!
  • Contents:
    • Compatibility
    • Usage instructions
  • Step 1: Subscribe to AutoML algorithm from AWS Marketplace
  • Step 2 : Set up environment
  • Step 3: Prepare and upload data
    • Background - The Iris dataset
  • Step 4: Train a model
  • Step 5: Deploy the model and perform a real-time inference
  • Step 6: Clean-up

ImplicitBPR

• Implementing a Recommender System for Implicit Feedback Datasets with Bayesian Personalized Ranking
  • Classes of recommender systems:
  • Explicit vs. Implicit Users Feedback:
  • About this Notebook
    • Step 1: Pre-requisites: subscribe to Implicit BPR Algorithm from AWS Marketplace
      • Notes:
    • Step 2: Set up environment
    • Step 3: Data collection and preparation
      • Step 3.1: Ingesting the Online Retail Dataset and load them inside pandas dataframe
      • Step 3.2: Exploring, cleansing and converting dataset into the format accepted by an algorithm
      • Step 3.3: Preparing the final training dataset and upload it to Amazon S3
    • Step 4: Train the model and evaluate the performance metrics
      • Step 4.1: Train the model
      • Step 4.2: Evaluate and visualize the performance metrics
    • Step 5: Perform a batch/offline inference
      • Step 5.1: Identify a customer and understand their purchase history
      • Step 5.2: Upload the payload to Amazon S3 and run a batch transform job
      • Step 5.3: Join the result with the stock lookup table to associate the item information
    • Step 6: Deploy the model and perform a real-time inference
      • Step 6.1: Deploy an endpoint
      • Step 6.2: Take the example user, create the JSON payload and make an inference request
      • Step 6.3: Join the result with the stock lookup table to associate the item information
    • Step 7: Cleaning up the Resources
  • Notebook CI Test Results

Use AWS Marketplace model packages

• AWS Marketplace Product Usage Demonstration - Model Packages
  • Using Model Package ARN with Amazon SageMaker APIs
  • Overall flow diagram
  • Compatibility
  • Set up the environment
    • Create the session
  • Create Model
  • Batch Transform Job
    • Batch Transform Input Preparation
      • Inspect the Batch Transform Output in S3
  • Live Inference Endpoint
    • Choose some data and use it for a prediction
    • Cleanup endpoint
  • Notebook CI Test Results

Auto insurance

• Goal: Automate Auto Insurance Claim Processing Using Pre-trained Models
  • Pre-requisites
  • Set up environment and view a sample image
  • Step 1: Deploy Vehicle Damage Inspection model
    • Step 1.1: Deploy the model for performing real-time inference.
    • Step 1.2: Perform a prediction on Amazon Sagemaker Endpoint created.
  • Step 2: Deploy the Vehicle recognition model.
    • Step 2.1: Deploy the model for performing real-time inference.
    • Step 2.2: Perform real-time inference on the model.
    • Step 2.3: Store the precise car image for further processing
  • Step 3: Extract labels from the picture (optional)
  • Step 4: View all outputs
  • Step 5: Cleanup
  • Notebook CI Test Results

Improving industrial workplace safety

• Demonstrating Industrial Workplace Safety using Pre-trained Machine Learning Models
  • Introduction
  • Pre-requisites
  • Step 1: Set up environment and view sample images
    • Step 1.1: Set up environment
    • Step 1.2: View sample images
      • Step 1.2.1: View construction site image
      • Step 1.2.2: View an image with a worker and a person at a workplace
      • Step 1.2.3: View an image with an excavator and a truck at work
  • Step 2: Deploy construction worker detection model
    • Step 2.1: Deploy the model for performing real-time inference.
    • Step 2.2: Perform a prediction (Test 1)
    • Step 2.3: Perform a prediction (Test 2)
  • Step 3: Deploy the hard-hat detection model.
    • Step 3.1: Deploy the model for performing real-time inference.
    • Step 3.2: Perform real-time inference on the model.
  • Step 4. Deploy the Personal Protective Equipment (PPE) detection model
    • Step 4.1: Deploy the model for performing real-time inference.
    • Step 4.2: Perform real-time inference on the model.
  • Step 5. Deploy the Construction Machines detection model
    • Step 5.1: Deploy the model for performing real-time inference.
    • Step 5.2: Perform real-time inference on the model (Test 1).
    • Step 5.3: Perform real-time inference on the model (Test 2).
  • Step 6. Generate actionable insights on video input
  • Step 7. Explore other relevant models!
  • Step 8. Cleanup
  • Notebook CI Test Results

Generic sample notebook

• Deploy and perform inference on ML Model packages from AWS Marketplace
  • Pre-requisites
  • Additional Resources
  • Contents
  • Usage instructions
  • 1. Subscribe to the model package
    • A. Identify compatible instance-type
    • B. Identify content_type
    • C. Specify model-package-arn
  • 2. Create an Endpoint and perform real-time inference.
    • A. Create an Endpoint
    • B. Create input payload
      • Identify a sample Input file you can use:
        • View the file you just downloaded
      • Video File
      • Audio File
        • Step B.1 Pre-process the data (Optional for some models)
      • Some models require Base64 encoded data
      • Some models require images in serialized format
      • Step B.2 Manually prepare data (applicable only if your payload is not ready yet)
      • Step B.3 Write payload you created to a file
    • C. Perform Real-time inference
    • D. Visualize output
    • E. Delete the endpoint
  • 3. Perform Batch inference
  • 4. Delete the model
  • 5. Cleanup
  • Notebook CI Test Results

GPT-2 XL use cases

• Creative writing using GPT-2 Text Generation
  • Overview
  • Contents
  • Usage instructions
  • Step 1: Determine input prompt and visualize word dependencies
    • Step 1.1 Introduction to attention
    • Step 1.2 Specify input prompt
    • Step 1.3 Visualize attention mechanism
  • Step 2: Use an ML model to generate text based on prompt
    • Step 2.1: Specify model arn from AWS Marketplace subscription
    • Step 2.2: Create model from model package and deploy to endpoint
  • Step 3: Explore use cases and model parameters
    • Step 3.1: Use case 1: Assisted writing of prose
    • Step 3.2: Use case 2: Autonomous authoring of poem
    • Step 3.3: Additional Use-Cases
    • Step 3.4: Delete Amazon SageMaker endpoint
  • Step 4: Use Amazon SageMaker batch transform
    • Step 4.1: Create input file for batch transform job
    • Step 4.2: Upload file to S3
    • Step 4.3: Execute the batch transform job
    • Step 4.4: Visualize output
    • Step 4.5: Delete the model
  • Step 5: Next steps
    • Step 5.1: Additional resources
    • Step 5.2: Cancel AWS Marketplace subscription
  • Notebook CI Test Results

Amazon Augmented AI (A2I) with AWS Marketplace ML models

• Amazon Augmented AI(A2I) Integrated with AWS Marketplace ML Models
  • Overview
  • Contents
    • Usage instructions
  • Prerequisites
    • Installing Dependencies
    • Setup Variables, Bucket and Paths
  • Step 1 Configure Amazon A2I service
    • Step 1.1 Creating human review Workteam or Workforce
    • Step 1.2 Create Human Task UI
    • Step 1.3 Create the Flow Definition
  • Step 2 Deploy and invoke AWS Marketplace model
    • Step 2.1 Create an Endpoint
    • Step 2.2 Create input payload
    • Step 2.3 Perform real-time inference
  • Step 3 Starting Human Loops
    • Step 3.1 View Task Results
  • Step 4 Next Steps
    • Step 4.1 Additional Resources
  • Step 5 Clean up resources
  • Notebook CI Test Results

Person Counting use case

• Evaluating ML models from AWS Marketplace for person counting use case
  • Introduction
    • Pre-requisites
      • Subscribe to the models
    • Download and Analyze dataset
      • Analyze data and annotations
    • Deploy SageMaker endpoints
    • Perform inferences
    • ML Model evaluation
      • Generic data evaluation
      • Few people in the frame evaluation:
      • Medium crowd evaluation:
      • Edge Cases
    • ML model evaluation summary
    • Cleanup
  • Notebook CI Test Results

Create synthetic features using pre-trained models in AWS Marketplace

• Supplement your data with synthetic features from a pre-trained model to train a better downstream audio classification model

Use AWS Marketplace open source model packages

PyTorch image classification models

• Deploy and perform inference on Model Package from AWS Marketplace
  • Pre-requisites:
  • Contents
  • Usage instructions
  • 1. Subscribe to the model package
  • 2. Create an endpoint and perform real-time inference
    • A. Create an endpoint
    • B. Prepare input file for performing real-time inference
    • C. Query endpoint that you have created with the opened images
    • D. Delete the endpoint
  • 3. Perform batch inference
  • 4. Clean-up
    • A. Delete the model
    • B. Unsubscribe to the listing (optional)
  • Notebook CI Test Results